Yeah, there's a feedback aspect to it: a more talented person will see greater results from practicing and thus be more encouraged to continue practicing. A less talented person will have a harder time and will be more inclined to give up.
Determining which genes are good for running is tricky, but to a limited extent since we can see and physically determine all the parts needed to run fast (skeletal, muscular, neurological, etc.--and beyond a basic level, self-control and bio-feedback). It's trickier of course to determine which genes are good for music because so much of it is exclusively in the mind. Obviously you need fast muscles in your fingers and wrists to be good at guitar, and a fast two-way connection between these and your eyes (if you're reading music), but it gets more nebulous when you get into stuff like creativity and "musical ear". There may not be a single set of genes that influences these things, but genes definitely have an impact. (We know this from adoption studies, twin-raised-apart studies, etc.) Just because we don't know exactly which genes are responsible for a given talent, or whether it's the same set of genes for that talent every time, doesn't mean the effect isn't real.
Also keep in mind that while genes are physical things that come in a limited number of combinations, the activity at which one might be talented is also nebulous and dynamic. A hundred years ago a good hockey player didn't have to worry about passing the puck forward, and he wasn't allowed to use a curved stick. Would a talented hockey player from 1916 transported into the future 100 years be talented today? Most likely, but he might find that a few of his talents don't quite align to the game the way it's currently played. In the same way, someone who's really talented at the accordion might never in a million years sit down at a computer and pump out a really great EDM song because the necessary skills for each only intersect so much.